formail and maildirs

dahamsta

I've just switched from mbox to maildirs, however I forgot about procmail. I
use this to mark messages read in my archive folders, what's the equivalent
for maildirs?

| formail -I "Status: R" >> Lists/$MATCH

adam

krinDar

Hi,

dahamsta wrote:

I forgot about procmail. I use this to mark messages read in my
archive folders, what's the equivalent for maildirs?

| formail -I "Status: R" >> Lists/$MATCH

adam

I assume you are doing something like:

:0:
* From: AMailList
| formail -I "Status: R"  >> Lists/$MATCH

And this no longer works because you have changed to Maildirs
and you cannot cat into a directory.

I believe this should do what you want.

:0:
* From: AMailList
{
    :0 fw
     | formail -I "Status: R"

     :0:
      Lists/MailList/
}

What this does is first determine if the mail is from a maillist.
If so it runs the nested commands. The first one, namely:

:0 fw
| formail -I "Status: R"

calls formail, treating it as a filter (flag 'f') and waits for formail to return
(flag 'w'). Formail changes the mail inline but does not redirect it to a
file, it hands it back to procmail.

The second piece just puts the mail into the Maildir for that list.

dahamsta

That's very helpful krinDar and I'm nearly there but I still have one more problem: The formail rule was used to mark these messages read, however apparently Maildir flags the filename rather than the headers to mark message status. Any idea how I'd go about that? (I'm googling away, just thought you might know offhand.)

Cheers,
adam

krinDar

dahamsta wrote:

That's very helpful krinDar and I'm nearly there but I still have one more problem: The formail rule was used to mark these messages read, however apparently Maildir flags the filename rather than the headers to mark message status. Any idea how I'd go about that? (I'm googling away, just thought you might know offhand.)

Sorry, I think I started to see only the trees and not pay attention
to the forest.

In each Maildir there are three directories:cur, tmp, new.
New, unread mail resides in the 'new' directory, mail that is in the process
of being delivered stays in 'tmp' until it is completed. Old, read mail resides
in the 'cur' directory.

To mark a mail read in a Maildir, you move it from the new directory
and append some information to the filename. This describes the
format of the filenames:
http://cr.yp.to/proto/maildir.html

Thus, I guess a hack to do what you want would be something like:

:0
* From.*mailist*
{
    :0
     lists/maildir

     :0
     | { cd lists/maildir; \
           for i in $(ls new)
           do
               mv new/$i cur/${i}:2,S
            done
        }
}

All this does it move the files and appends the information to mark
the file/mail read as necessary.

This is a Kludge IMHO, and I have not tested it.

You could also try to deliver to the Maildir _file_ as opposed to the
Maildir itself. Typically we tell procmail to deliver to lists/maildir,
and procmail handles to lower level problems, which
directory it should put the mail into (lists/maildir/new) and the file
it should put the mail into (as describe above).

You could do this work yourself and save the mail to the
lists/maildir/cur directory, e.g

# replace with your hostname
# this is the Maildir file to save to
file=`perl -e 'printf("%s.%s_1.HOSTNAME:2,S",time,$$)

:0:
* From.*Mailist*
lists/maildir/cur/$file

I have tested this quickly and it appears to work a treat, it is
however a bit of a kludge and could cause you problems depending
on how you run procmail, if you get more than one mail for that
list when you run procmail then we will try and put it into the
same file, , though I do not believe this would be a
problem, AFAICR procmail is called for each mail that is being
delivered, so the file will be generated each time, with a new
process number and possibly a new time, thus file should be
unique.

dahamsta

First one's a bit top-heavy since there'll be a substantial amount of mail in these directories, but the second solution is very trick, would seem to be exactly what I'm looking for. I'll set up a receipe with an extra backup and try it out, thanks.

adam

krinDar

dahamsta wrote:

First one's a bit top-heavy since there'll be a
substantial amount of mail in these directories,

Yeah it is a really rough approach that is bound to be affected by external
things.

but the second solution is very trick, would seem to be exactly
what I'm looking for. I'll set up a receipe with an extra backup and try
it out, thanks.

Hopefully it will do what you want. What I could see going wrong with it, is that
because you are delivering to an actual file, procmail will treat it as a mbox,
thus it might manage to append multiple messages to the one file. You probably
would not loose mail, just your maildir would be full of mboxes.

The way this could occur is if procmail gets called with a batch of
messages, as the variables would not change for that batch of
messages. If you call procmail through your .forward file, this should
not be a problem.

A potential way to get around this is to increment the variable which
counts the number of messages delivered in that batch. i.e in this
file: 103342534.9435_C.moo.spoo.org:2,S, C is the counter. This
is defined by the spec for Maildir so it is not that strange to do so.

How to increment a variable within an action/rule, I am not sure.

I guess another method that would be much more failsafe, but perhaps
a little slower would be to write a utility to look in the maildir provided as
an argument, create a unique filename, and put the mail in there.

Do let us know how you get one with it.

dahamsta

It's a little silly that procmail doesn't support this kind of thing by default, it's not exactly an unusual request. What bugs me is that when I asked my question on the Courier list it went ignored, and it's not the first time this happened either: my post included a reference to a previous unanswered question. I hate it when people's answer is to simply ignore the question, it's plain ignorant.

Anyway, I'm nearly there now, I just have one more issue: I'm sorting this mail on the fly, from various Mailman mailing lists, however my test messages are failing for the simple reason that the directory for the list doesn't exist. This will happen in practise too as I subscribe to new lists, so I'm wondering what would be the best way of creating the directory on the fly?

I know in the shell it would be something like if [ -e $DIR/ ], but I'm not sure how to integrate that into my procmail file. Here's my recipe so far:

FILE=`perl -e 'printf("%s.%s_1.tedcrilly:2,S",time,$$)`

:0
* ^X-BeenThere: .+$
{
   :0:
   .Lists/

   :0 fw
   | formail -I "Status: R"

   :0:
   * ^X-BeenThere: \/.+
   .Lists/$MATCH/cur/$FILE

}

adam

krinDar

dahamsta wrote:

It's a little silly that procmail doesn't support this kind of thing by default, it's not exactly an unusual request.

Creation of directories or writing directly to Maildir style files ?

I am not sure, I mean, if you start adding loads of features like that
I think it would become a little bloated. It is amazingly powerful to be
able to call external "programs" however.

I hate it when people's answer is to simply ignore the question, it's plain ignorant.

*Ahem* - Sorry it took me so long to respond.

so I'm wondering what would be the best way of creating the directory on the fly?

I know in the shell it would be something like if [ -e $DIR/ ], but I'm not sure how to integrate that into my procmail file.

In your nested rules you should be able to add the following:

:0
| if [ ! -e $DIR ]; \
   then mkdir -p $DIR; fi

I think this is probably the best way of achieving what you want to do.